Is Translation Technology Making Learning Languages Irrelevant?
Facebook recently announced the acquisition of Mobile Technologies, known for its speech-to-speech translator product called Jibbigo. Google is also researching to come up with a real-time translation device to help customers interact with people speaking different languages. Earlier, Microsoft too demonstrated a universal translator that not only translates what you are saying but also sounds just like you while doing so.
Translation technology can be broadly categorized as text-to-text translation and speech-to-speech translation. Text-to-text translation is also referred to as machine translation. Though not completely accurate, it has gained some good maturity and production use in past few decades.
Speech-to-speech recognition and translation is perceived as an area of immense opportunity that can potentially simplify our lives in various areas. In its current form, some of the banking systems, applications like Microsoft Word, and Siri use this technology to perform basic tasks. At a high level, Microsoft's universal translator and probably others too operate in three basic steps:
- Convert spoken source language word-by-word to the target language
- Arrange the word order of resulting sentence to suit target language
- Convert the last output back to speech
With the advancements in machine translation, the first step above can be rendered with fair accuracy (though not 100 percent). It is observed that the biggest challenges are recognizing the differences in syntax—the ordering of words across languages—as different languages have different word orders.
The accent and delivery of speech also varies from person to person, which makes it harder to form standard patterns to recognize and translate speech. Other factors, such as dependency on hardware like microphones and environmental noise, make it more challenging to deliver a flawless output.
Despite these constraints and with advancements in technology, Microsoft's technology has been observing reduction in the word error rate for speech by more than 30 percent compared to previous methods. Google's device is reported to be working fairly accurately on some languages, such as English to Portuguese.
Arguably, the overuse of translation technology may widen the gaps between world cultures when people do not make an effort to learn a language. And there are numerous studies that prove that learning a new language is good for brain development.
By Google's admission, the interpreter device is still years away from being a reality. Rick Rashid, Microsoft’s Chief Research Officer, presents an optimistic view, "...we may not have to wait until the twenty-second century for a usable equivalent of Star Trek’s universal translator..."
It is evident that it may take several years—or even decades—for the translation technology to mature enough to be in a position to replace the human's need of learning languages. Until that happens, this journey is surely an exciting one to watch and follow.
